bug#40323: 28.0.50; error in process filter: Invalid search bound (wrong side of point)

[Jacob Lagares Pozo]

Hello Noam,

I'm sorry for the late reply. I have been having some problems with the power source of my computer as of lately. Everything seems to be doing fine now though, so I'll get back to this.

I made a very simple program that just prints a bunch of stuff to stdout and stderr:

If I run it without your patches, it works surprisingly just fine (I noticed the original errors pop up most commonly on Slack I guess because it prints a lot more?), whereas if I evaluate said patches, this is the output of the trace buffer:

======================================================================
      1 -> (comint-output-filter #<process Shell> "stdout:
      hello, world
      Im gonna print some stuff
      0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
      stderr:
      AHHHH PANIC CATASTROPHIC FAILIURE!!!
      The quick brown fox jumps over the lazy dog.
      ")(:comint-pmark nil)
      | 2 -> (set-marker #<marker in no buffer>
      1)(:comint-pmark nil)
      | 2 <- set-marker: #<marker at 1 in *Async Shell
      Command*>(:comint-pmark (#<marker at 1 in *Async Shell
      Command*> . #<marker at 1 in *Async Shell Command*>))
      | 2 -> (set-marker #<marker at 1 in *Async Shell
      Command*> 191)(:comint-pmark (#<marker at 1 in *Async Shell
      Command*> . #<marker at 1 in *Async Shell Command*>))
      | 2 <- set-marker: #<marker at 191 in *Async Shell
      Command*>(:comint-pmark (#<marker at 1 in *Async Shell
      Command*> . #<marker at 191 in *Async Shell Command*>))
      | 2 -> (ansi-color-process-output "stdout:
      hello, world
      Im gonna print some stuff
      0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
      stderr:
      AHHHH PANIC CATASTROPHIC FAILIURE!!!
      The quick brown fox jumps over the lazy dog.
      ")(:comint-pmark (#<marker at 1 in *Async Shell Command*> .
      #<marker at 191 in *Async Shell Command*>))
      | 2 <- ansi-color-process-output: nil(:comint-pmark
      (#<marker at 1 in *Async Shell Command*> . #<marker at
      191 in *Async Shell Command*>))
      | 2 -> (comint-adjust-window-point #<window 31 on *Async
      Shell Command*> #<process Shell>)(:comint-pmark
      (#<marker at 1 in *Async Shell Command*> . #<marker at
      191 in *Async Shell Command*>))
      | 2 <- comint-adjust-window-point: nil(:comint-pmark
      (#<marker at 1 in *Async Shell Command*> . #<marker at
      191 in *Async Shell Command*>))
      | 2 -> (set-marker #<marker (moves after insertion) at 191
      in *Async Shell Command*> 191)(:comint-pmark (#<marker at 1
      in *Async Shell Command*> . #<marker at 191 in *Async Shell
      Command*>))
      | 2 <- set-marker: #<marker (moves after insertion) at 191
      in *Async Shell Command*>(:comint-pmark (#<marker at 1 in
      *Async Shell Command*> . #<marker at 191 in *Async Shell
      Command*>))
      1 <- comint-output-filter: #<marker (moves after insertion)
      at 191 in *Async Shell Command*>(:comint-pmark nil)

I am not sure what does this mean, perhaps it is some special character Slack uses for logging that messes with those markers, I don't know. Maybe I could try printing all of the ASCII characters sequentially and see what happens.

Regardless, I'm still not entirely sure what your code is doing anyway.

Thanks, Jacob

PS: I don't know if I might have accidentally sent the message halfway writing it; if you see anything weird in another message, that might be the reason why.

On 21/04/2020 04:29, Noam Postavsky wrote:

> Jacob Lagares Pozo <jlagarespo@iebesalu.cat> writes:
>
> > I should probably make a simple program that prints a bunch of stuff
> > and then hangs, so I can have predictable and reproducible output,
> > that might help.
> >       
> It occurs to me that you should see a "non-local exit" in the trace when
> the error triggers, and the traces just before that should hopefully
> show the swapping of marker positions occuring.
>
> > So what do you exactly mean by that the process is ending normally?
> >       
> Oh, hmm, I was still a bit confused.  I thought the (:comint-pmark nil)
> meant the marker was deleted, but actually it's just because around the
> call to comint-output-filter a different buffer is current (which makes
> the check in the tracing function fail).  Maybe one more tweak to the
> tracing function:
>
>     (defun bug-40323-get-comint-output-marker ()
>       (list :comint-pmark
>             (let ((buf (and (markerp comint-last-output-start)
>                             (marker-buffer comint-last-output-start))))
>               (when (buffer-live-p buf)
>                 (cons
>                  comint-last-output-start
>                  (process-mark (get-buffer-process buf)))))))
>